podserver response times

Pod::Parser performance (therefore also Pod::Server) is disappointingly
slow when handling medium size (5k-50k) files. Memory usage climbs rapidly
because it seems Parrot or Rakudo leaks memory until the process exits.

The operating system resorts to swapping hectically in a hopeless effort
to provide more, but Parrot doesn't recycle and keeps calling malloc().
After almost hanging the computer for a long time the kernel usually kills
the Parrot.

Running L<Pod::Server|http://github.com/eric256/perl6-examples/blob/master/lib/Pod/Server.pm>
and L<Pod::Parser|http://github.com/eric256/perl6-examples/blob/master/lib/Pod/Parser.pm>
causes huge memory and CPU loads. Some approximate Pod::Server responses:

 Document                                   size  time   RAM
 rakudo/docs/compiler_overview.pod            9k   54s  224M
 rakudo/docs/glossary.pod                     4k   19s  132M
 rakudo/docs/guide_to_setting.pod             3k   15s  114M
 rakudo/docs/release_guide.pod                1k    5s   94M
 pugs/docs/Perl6/Spec/S01-overview.pod        5k   28s  158M
 pugs/docs/Perl6/Spec/S02-bits.pod          151k  600s 1583M (died)
 pugs/docs/Perl6/Spec/S03-operators.pod     150k
 pugs/docs/Perl6/Spec/S04-control.pod        58k  340s 1048M (died)
 pugs/docs/Perl6/Spec/S05-regex.pod         133k
 pugs/docs/Perl6/Spec/S06-routine.pod       113k
 pugs/docs/Perl6/Spec/S07-iterators.pod       7k   35s  174M
 pugs/docs/Perl6/Spec/S09-data.pod           46k  255s  820M (died)
 pugs/docs/Perl6/Spec/S10-packages.pod        8k   49s  204M
 pugs/docs/Perl6/Spec/S10-modules.pod        22k  137s  433M

Results on a dual core amd64 4200+, 2GB memory, running Linux. Only one
CPU core runs each process :(

Seeing memory use up to 15000 times document size is worrying.
The developer guidelines that apply to Java and .NET probably apply to
Rakudo as well - keep the application code small and shift as much of
the heavy processing burden as possible to the standard libraries of the
language.